Development of an Expert System for the Optimal Design of Roof and Bridge Trusses

Ojoko, N.K (2012-11)

A Thesis Submitted to the School of Postgraduate Studies, University of Lagos

Thesis

Several attempts have been made to optimize the design of bridge and roof trusses. This work presents the development of an Expert System (ES) for the optimal design of bridge trusses subjected to moving loads and roof trusses as well. The research improves on such previous attempts in many areas including automation of structure data generation, experience gaining facilities, suggestions and decision making, flexibility, and structural arrangement, shortest possible volume of input data, ability to carry out self improvement, and dependability for professional adoption. Least-weight truss optimization which translates to least-cost design was achieved for the truss configuration for any given span of trussed bridge or roof truss. The objective function necessary for minimal weight of structure subject to serviceability constraints was derived. Five classes of bridge and roof trusses considered are the Warren, braced Warren, N-, Pratt and braced Pratt trusses. The bridge trusses are arranged to take loads from the bridge deck through stringer and floor beams to the joints. The expert system was encoded based on appropriate design codes and rules of thumb garnered from experienced professional engineers. The system initializes its knowledge base for spans and loads of sufficiently wide ranges in both cases. The objects of the database include the optimum truss types stored in memory with coded numbers to maximize storage memory. The knowledge base expands with use, which makes the expert system to become more intelligent. Evaluation of the performance of the developed system demonstrates high accuracy, efficiency, savings in data entry, computational time and ease of reuse. The compiled form of the resulting product is useable on all forms of digital computers. Comparison of its results with those of popular industrial software demonstrates its superior performance in all cases.

Collections: